Validity and reliability of the DANU sports system for walking and running gait assessment

Objective. Gait assessments have traditionally been analysed in laboratory settings, but this may not reflect natural gait. Wearable technology may offer an alternative due to its versatility. The purpose of the study was to establish the validity and reliability of temporal gait outcomes calculated by the DANU sports system, against a 3D motion capture reference system. Approach. Forty-one healthy adults (26 M, 15 F, age 36.4 ± 11.8 years) completed a series of overground walking and jogging trials and 60 s treadmill walking and running trials at various speeds (8–14 km hr−1), participants returned for a second testing session to repeat the same testing. Main results. For validity, 1406 steps and 613 trials during overground and across all treadmill trials were analysed respectively. Temporal outcomes generated by the DANU sports system included ground contact time, swing time and stride time all demonstrated excellent agreement compared to the laboratory reference (intraclass correlation coefficient (ICC) > 0.900), aside from ground contact time during overground jogging which had good agreement (ICC = 0.778). For reliability, 666 overground and 511 treadmill trials across all speeds were examined. Test re-test agreement was excellent for all outcomes across treadmill trials (ICC > 0.900), except for swing time during treadmill walking which had good agreement (ICC = 0.886). Overground trials demonstrated moderate to good test re-test agreement (ICC = 0.672–0.750), which may be due to inherent variability of self-selected (rather than treadmill set) pacing between sessions. Significance. Overall, this study showed that temporal gait outcomes from the DANU Sports System had good to excellent validity and moderate to excellent reliability in healthy adults compared to an established laboratory reference.


Introduction
Gait analysis involves the systematic study of human walking or running, whereby quantitative information on walking or running performance and abnormalities arising from musculoskeletal (Bramah et al 2018), cardiopulmonary (Zago et al 2018, Liu et al 2019) and neurological pathologies (Celik et al 2021) or injuries can be obtained (Dever et al 2022).As a result, gait analysis has been employed in sports performance and medicine, where information can be used to improve athlete performance (Boulgouris et al 2005, Shun-Ping et al 2014, Moore 2016, Burns et al 2019, Mason et al 2022) or diagnose and monitor injury or health conditions (Meardon et al 2011, Noehren et al 2012, Baker et al 2016).For example, gait has been found to be a useful biomarker for neurological concussion injuries (Celik et al 2021, Powell et al 2021, Dever et al 2022) or musculoskeletal injuries (Bramah et al 2018), which would otherwise be undetected.
Traditional gait analysis has largely been performed in a laboratory setting using 2D video analysis (Pipkin et al 2016, Dingenen et al 2018), 3D motion-capture systems (Pfister et al 2014), force plates (Leitch et al 2011), instrumented walkway mats or treadmills (Donath et al 2016, Higginson 2009, Parati et al 2022).Although these methods have high accuracy in measuring gait outcomes (Dugan andBhat 2005, Higginson 2009), there are inherent drawbacks, such as the expense of equipment, the need for trained practitioners to collect and analyze data and the requirement to attend a laboratory setting.Therefore, traditional 'gold-standard' techniques are not readily available within sport performance or clinical settings, and they lack generalizability and ecological validity (Dugan andBhat 2005, Higginson 2009) (i.e.gait in a laboratory may not reflect gait in the real-world).Laboratory settings lead to the use of constrained protocols that may not represent typical real-world gait, such as assessing intermittent trials of single foot strikes on force plates, with unnatural force platform targeting (Challis 2001) and limited numbers of consecutive steps (Higginson 2009), whereas runners approximately take 1500 steps per mile (Hoeger et al 2008) and the clinical and general populations typically take 5000-10 000 steps per day in the real-world (Schuna et al 2013, Del Pozo Cruz et al 2022, Lempriere 2022).Gait analysis algorithms typically perform best under continuous walking bouts and with greater duration of recording, and due to potential changes in mechanics over long periods of walking or running, analyzing an abundance of steps may be beneficial (Storm et al 2018, Toth et al 2023, Veerubhotla et al 2021).Numerous studies have sought to overcome the issue of intermittent overground trials by using instrumented treadmills, however, further studies demonstrate the inconsistencies in gait between over-ground and treadmill locomotion (Chambon et al 2015), with the treadmill providing an external cue for gait (i.e.external rhythm of gait is set by the treadmill which influences gait metrics) (Thumm et al 2018).In order to enhance understanding of real-world gait, a range of intermittent and continuous gait tasks or conditions may be required within laboratory gait analysis assessment to represent the spectrum of gait (and gait outcomes) seen in the real-world (Mann et al 2016).
Wearable technology offers a low-cost (affordable) and lightweight alternative to overcome traditional gait assessment limitations (Stuart et al 2021), with such technology becoming increasingly accepted and adopted by users (sports professionals, patients etc) and clinicians (Willy 2018).The majority of commercial or researchgrade wearables that have previously been used for gait analysis include accelerometers, gyroscopes, and magnetometers applied individually or in combination as an inertial measurement unit (IMU) (Tao et al 2012, Mason et al 2022).More recently, advances in textile technology have allowed for development of multi-modal devices through integration of pressure sensors and IMUs into flexible material that can be continuously worn in an unobstructive manner to provide comprehensive gait outcomes within any environment, such as instrumented socks.While wearable technologies for gait assessment are being increasingly used, fewer than 10% of commercially available wearable technologies for gait analysis are analytically validated against an accepted 'gold-standard' (reference tool) (Storm et al 2016), with even fewer establishing reliability of wearable sensor derived gait outcomes (Mason et al 2022).Establishing the analytical validity and reliability of such wearable technologies against reference tools is vital to ensure that underlying algorithms that provide gait outcomes are accurate and provide reliable outcomes within specific populations (i.e.healthy or clinical groups) that performance and clinical decisions could be definitively made (Goldsack et al 2020, Rochester et al 2020).Following initial analytical validation within cohorts of interest, wearable devices and outcomes can be examined for specific performance or clinical use with investigation of gait in various settings (i.e.lab or outdoor/realworld), thus developing greater understanding of gait in both clinical (i.e.neurological, musculoskeletal, or cardio-pulmonary conditions) (Hulleck et al 2022) and sporting contexts (i.e.performance, fatigue, and injury mechanisms) (Stuart et al 2021).
The DANU Sports System (DANU, Ltd, Dublin, Ireland) is a commercially available system that combines capacitive pressure sensors on the sole of the foot and tibia based IMUs (one on each leg) encompassed within a sock format, which wirelessly synchronizes and streams data to a mobile tablet.The DANU System offers a large quantity of capacitors and high sampling frequency (i.e. 15 capacitors and 250 Hz in DANU System) that is greater than other similar wearable systems (e.g. 13 capacitors and 50 Hz, Moticon insoles (Moticon GmbH, Munich, Germany)) (Stöggl and Martiner 2017), which will allow for more comprehensive and potentially more accurate gait outcomes (Tao et al 2019).The DANU system software allows for collection of gait data from the wearable socks within any environment and provides automatic analysis, without the requirement for gait research knowledge or expert data processing.The DANU software package allows the user to select various instrumented tests of walking or running (intermittent or continuous tasks) and automatically generates a report for each participant (or group of participants).At present, the DANU report provides temporal gait outcomes derived from initial contact (heel strike) and final contact (toe-off) of the feet, which include ground contact time (GCT), stride time and swing time.Monitoring temporal gait characteristics is important in clinical and sporting contexts.Regarding sport performance, GCT is the most reported outcome in running wearables and has been indicated as a critical factor to running economy (Santos-Concejero et al 2015, Mason et al 2022).Morin et al (Morin et al 2007) demonstrated that 90%-96% of the variance in leg stiffness can be explained by GCT, in turn less economical runners are shown to have a more slacken running style during ground contact as reflected by the low vertical stiffness (Heise andMartin 2001, Moore 2016).In clinical settings, temporal gait outcomes have been shown to be useful clinical measures that can detect gait deterioration due to aging (Hollman et al 2011, Chung et al 2022), or pathology (Lemke et al 2000) and improvements in gait due to rehabilitation or training (Patterson et al 2008, Smania et al 2011, Vitale et al 2012, Abd El-Kafy and El-Basatiny 2014).The simple and automatic gait output from the DANU system ensures that data collection is accessible and interpretable within performance or clinical settings, or within large trials.However, the DANU commercial multi-modal system has yet to be evaluated for analytical validity or reliability of gait outcomes, and such systems require validation against robust previously validated systems ('gold-standard reference tools') in controlled environments (i.e.laboratories) before being further validated and deployed within performance or clinical settings, or clinical trials (Goldsack et al 2020, Rochester et al 2020).Therefore, the purpose of this study was to determine the analytical validity, as well as test re-test reliability of gait outcomes obtained via the DANU Sport System during walking and running compared to a concurrently used 'gold-standard' 3D motion capture system in healthy adults.

Participants
Forty-one healthy participants were recruited from running clubs in the North-East of England.Inclusion criteria required participants to be aged >18 years, able to run unassisted for short periods.Prior to testing, all participants completed a questionnaire to provide information pertaining to their demographics, injury, and medical history, sporting pursuits and running personal bests.Injury was classified as 'any muscle, bone, tendon or ligament pain in the lower back/legs/knee/foot/ankle that caused a restriction or stoppage of running (distance, speed, duration or training) for at least 7 d or 3 consecutive scheduled training sessions, or that required the runner to consult a physician or other health professional' (Yamato et al 2015).Ethical approval was granted by the Northumbria University Research Ethics Committee (reference: 33358) and this study conformed to the Declaration of Helsinki.All participants were supplied with informed consent and gave verbal and written consent before performing testing in Northumbria University's Gait and Biomechanics Laboratory, City Campus.

Wearable technology: DANU sports system
The DANU sports system (figure 1) consists of a pair of textile socks, that were worn on both feet.Each sock contains 15 silicone based capacitive pressure sensors, and an IMU module that attaches to the medial surface of the mid-shank of tibia.Each IMU module is Bluetooth enabled for data transmission and is comprised of two configurable tri-axial accelerometers (Accelerometer 1 ±2 g, ±4 g, ±8 g or ±16 g, Accelerometer 2 ±100 g, ±200 g and ±400 g), gyroscope (±2000°s −1 ), magnetometer, and with variable sampling rates (60-250 Hz).The IMU module includes in-built memory for data collection.Here, the DANU sports system was configured to a default sampling rate of 250 Hz, ±16 g, ±200 g accelerometers and ±2000°s −1 gyroscope.A standing calibration trial was recorded prior to participant assessment.Data was collected via Bluetooth on Apple devices 2018 or later (iPad or iPhone devices are required to have at least Bluetooth 5.0 connectivity), and data processing was run through a custom-made Apple application for real-time feedback and visualization, as per the manufacturer's guidelines.

Reference systems
During the overground trials the reference system consisted of a 14-camera 3D motion capture system, distributed around a space of 9.8 × 7.9 × 3.2 m 3 , sampling at 250 Hz (VICON, Oxford, UK) and two staggered 0.5 m-long force plates (AMTI, Watertown, MA, USA), sampling at 1000 Hz, embedded in the middle of a walkway.For the treadmill running trials the reference system consisted solely of the 3D motion capture system.The calibration of the Vicon system was conducted before each participant.Sixteen reflective markers were placed on the participants lower limb before testing, and a static calibration trial was initially collected to form a musculoskeletal model (Kim et al 2021).Using a small amount of double-sided tape, the markers were attached bilaterally to the following landmarks: anterior superior iliac spine, posterior superior iliac spine, mid-lateral thigh, lateral knee joint line, lateral mid-shank, lateral malleoli, calcaneal tuberosity, and base of the second metatarsal.
Participant specific information of weight, height, ankle width, knee width, and leg length (from posterior iliac spine to medial malleolus) were measured and inputted in the lower body model (Sabharwal and Kumar 2008).The Plug-in-Gait (PiG) lower body model was used to analyse movement at the joints and evaluate all parameters (Leboeuf et al 2019).The lower body was modelled as seven segments (one pelvis, two thighs, two shanks, and two feet).A normal gait cycle was defined from the initial heel-to-heel contact with the same limb.Additional information of the PiG calculations can be found on Vicon's website.
Data processing was performed in Vicon Nexus.All markers were labelled, and marker trajectories were filtered using a fourth order low-pass Butterworth filter via dynamic plug-in gait model with 6 Hz cut-off frequency.For the overground trials, identification of gait events (initial contact and toe-off) was determined using the vertical ground reaction force from force plate data from Vicon Nexus.These events were detected by applying a threshold of 20 N in vertical ground reaction force (Smith et al 2015).For treadmill trials, gait identification was achieved through visual inspection of initial contact and toe off for consecutive strides over the trials.The trajectory of the heel and toe markers in the Z plane were examined, so that the minimum of the trajectory of one stride specified the timestamp of initial contact.The traced trajectory of the toe marker was used to specify the toe's movement, so that the minimum of the trajectory specified the timestamp of a toe-off event.The initial contact and toe-off events of left and right foot steps were combined in order to estimate for each step GCT, swing time and stride time (Falbriard et al 2018).Ground contact time and swing time were defined by the time between initial contact and toe-off events and between toe-off and initial contact events, respectively.

Procedures
A concurrent validation study was conducted to determine agreement between the DANU sports system and the 3D motion capture system (Mason et al 2023).Prior to commencing the protocol participants were provided the opportunity to run on the treadmill (Spirit fitness XT485) at a comfortable speed for a warm-up and to familiarise themselves.For the overground trials, participants were asked to walk at a self-selected speed across the walkway (10 m), three trials were collected.This process was repeated for the over-ground running trials.For the treadmill trials, participants completed 60 s of walking at a self-selected speed and then ran at four standardised speeds (i.e. 8, 10, 12 and 14 km hr −1 ).If a participant could not reach a certain speed (i.e. 12 or 14 km hr −1 ) or did not feel comfortable at that speed, then it was not completed (See table 1).To ensure participant safety, the order of speed was consistent across participants, starting at the slowest speed (i.e. 8 km hr −1 ) and progressing to the fastest (i.e. 14 km hr −1 ).Data was collected for 60 s at each speed.A period of 60 s was chosen as it generally aligns with other similar studies in the field with data capture periods ranging from 20 s (McGrath et al 2012) to 90 s (Bailey and Harle 2015, Tan et al 2020, Mason et al 2022).Participants could have breaks between trials or could abort the trial at any time.Participants were provided with a standardised, neutral cushioning running shoe (Saucony Guide Runner) to wear during testing to ensure consistency and remove bias from gait-affecting cushioning within e.g.support cushioning running shoes (Roca-Dols et al 2018).The reference and wearable technologies were recorded simultaneously to allow direct comparison of the gait outcomes.To assess test re-test reliability, participants completed the protocol in the same format in a repeatedmeasures design, approximately one week after the first session.
The outcome measures were the temporal gait characteristics as measured by the DANU sports system and reference system, GCT, swing time and stride time.Outcomes were derived from the proprietary DANU Sport System gait algorithms that processed data within the DANU software and mobile application/cloud.In brief  outcomes were defined as follows: GCT was measured as the time (in ms) elapsed between initial contact (where the foot first contacts the ground) and final/terminal contact (where the foot last leaves the ground).Swing time was defined as the time (in ms) the foot spends off the ground in the gait cycle, defined by the time from toe off to heel strike of the same foot.Stride time was measured as the time (in ms) between two consecutive heel strikes of the same foot time.These outcomes were obtained across overground and treadmill trials, during walking and running.Outcomes were averaged over the one-minute trials.

Statistical analyses
All gait outcomes calculated by DANU sports system were comparatively analysed with the same outcomes calculated within the 3D motion capture data.Gait outcomes were averaged over the 60 s trials.Data analysis was conducted in SPSS v27 (SPSS Inc., Chicago, IL, USA).Shapiro-Wilks tests indicated a normal distribution of all data (p < 0.05).Subsequently, intra-class correlation (IC(2,1)) models examined absolute agreement between the DANU sports system and the reference (Zago et al 2018)D motion capture system.A predefined ICC performance scale was deployed, defined as poor (<0.50), moderate (0.50-0.75), good (0.75-0.90) or excellent (>0.90) (Koo and Li 2016).Mean error were calculated between the DANU sports system and the reference (Zago et al 2018)D motion capture data for descriptive purposes and are observed as an accuracy metric in the outcomes.In order to demonstrate the bias within the limits of agreement (LoA) were calculate and Bland-Altman plots were used to visually assess the agreement between systems (Bland and Altman 1986).To determine the test re-test reliability of the DANU sports system, Pearson's correlation coefficients (r), ICCs and LoA between the two testing time-points were calculated (Shrout and Fleiss 1979).An acceptable statistically significant threshold was set at p < 0.05.

Participants
A total of forty-one participants completed the study (26 Male, 15 Female; 36.4 ± 11.8 years; 173.3 ± 8.7 cm; 72.6 ± 12.2 kg).Participants exhibited a range of running abilities (5 km personal best; 23:31 ± 04:49).Of the 41 participants, some data loss or dropout during higher speeds was experienced (table 1).Upon preliminary observation of the quantified outcomes, no significant outliers were identified.For the validation aspect, a total of 1406 steps during overground walking and jogging trials were analysed.For treadmill testing, a total of 613 trials across all speeds were examined.For reliability analysis, 666 overground trials and 511 treadmill trials across all speeds were examined.Table 1 shows the descriptive gait data statistics from, along with the absolute agreement between the two systems for ICC, LoA (% and 95%) and r values.The agreement between the DANU sports system and 3D motion capture is visually displayed via Bland-Altman plots in figure 2.

Swing time
Intraclass correlations demonstrate an excellent agreement between the DANU sports system and the reference system for swing time (ICC(2,1) >0.900, LoA% 3.9 to 9.3).Robustness at the full range of speeds is demonstrated, with low mean differences throughout (1.06-9.26ms).The DANU sports system tends to over-    (b).With respect to sex and validity of the DANU sports system for measuring swing time, no significant differences were found (supplementary tables 1(a) and 1(b)).Excellent reliability (ICC(2,1) >0.900, LoA% 3.0 to 4.6) across all treadmill running trials can be seen, with good reliability during treadmill walking (ICC(2,1) 0.886, LoA% 9.0).Overground walking and jogging demonstrated moderate reliability (ICC(2,1) 0.698 and 0.672, respectively, LoA% 9.3 and 5.1), table 2. Excellent reliability for swing time during overground jogging was observed for females (ICC(2,1) 0.941, LoA% 15.9), compared to moderate reliability in males (ICC(2,1) 0.677, LoA% 17.6).No additional significant differences were observed in the reliability of the DANU sports system for measuring swing time with respect to sex (supplementary tables 2(a) and 2(b)).

Stride time
Intraclass correlations demonstrate an excellent agreement between the DANU sports system and reference system for stride time across all trials (ICC(2,1) >0.900), with low mean differences throughout (0.45-4.04 ms), table 1 and figures 2(c) and 3 (c).With respect to sex and validity of the DANU sports system for measuring stride time, no significant differences were found (supplementary tables 1(a) and 1(b)).

Discussion
The present study conducted an examination of the analytical validity and test re-test reliability of gait outcomes (temporal outcomes of GCT, stride time and swing time) measured by the DANU sports system, demonstrating that walking and running gait outcomes had good to excellent agreement with a 'gold-standard' reference and moderate to excellent reproducibility in healthy adults in laboratory conditions.Across walking and running trials, the identified gait outcomes and differences between the chosen gait speeds fall within the expected ranges when compared to other validated and established gait analysis systems in healthy adults (Lee and Hidler 2008, Braun et al 2015, Mason et al 2022), and reliability values are consistent with those reported for other wearable technologies for gait analysis (Godfrey et al 2014).

Validity and reliability of gait outcomes during walking
Temporal gait outcomes of GCT, stride time and swing time that were derived during overground and treadmill walking in healthy adults, which generally had excellent validity (ICC(2,1) >0.900) (tables 1 and 2, figures 2 and 3).The test re-test reliability results indicated a moderate to good reliability (ICC(2, 1) 0.698-0.750)during overground walking trials and moderate to excellent reliability (ICC(2, 1) 0.886-0.9893)during treadmill walking trials.These analytical validation results are important, as temporal walking gait outcomes are clinically relevant/meaningful metrics.For example, temporal metrics are sensitive to classify fallers and non-fallers in neurological patients (Zhou et al 2020, Shema-Shiratzky et al 2022).
Our findings are comparable to others that have validated or examined repeatability of wearable technology for gait analysis.Other systems commonly detect gait by placing sensors, typically IMUs or accelerometers, at the lumbar (Bugané et al 2012, Morris et al 2019), tibia (Iosa et al 2016, Mancini andHorak 2016), or shoe (Donath et al) of the participant.Many other systems are shown to be reasonably valid and reliable (Henriksen et al 2004, Iosa et al 2016).Yet, this is the first study to use a gait analysis system that combined capacitive pressure sensors on the sole of the foot and tibia based IMUs encompassed within a sock form factor.In contrast to the current study, previous work has demonstrated better validity and reliability for basic spatiotemporal gait outcomes, such as GCT or stride time, rather than outcomes of relative phase, such as swing time that show poorer agreement (Aminian et al 2002, Sabatini et al 2005).Difficulties detecting relative phases of the gait cycle with other wearable systems have been suggested to be due to limitations in the accuracy of detecting toe off events (Washabaugh et al 2017), which may also underlie poorer agreement for the DANU systems GCT outcome.Typically, the outcomes of interest that are used for gait analyses are spatiotemporal gait outcomes which require the identification of initial contact (i.e.heel strike) and toe off events for each step (Sprager andJuric 2015, Benson et al 2019).However, the DANU systems gait event detection algorithm is proprietary, which limits detailed discussion or understanding of agreement results, which may require algorithms to become opensource to allow for future improvement.
The moderate reliability of temporal gait outcomes when walking may be due to the task being undertaken, as overground and treadmill walking have different underlying muscle activity (Lee and Hidler 2008).Additionally, self-selected paced overground trials may lead to unnatural force plate targeting (Challis 2001) and are influenced by landing patterns when compared with natural gait (Van Hooren et al 2020).The aforementioned could shed light on why variations in reliability emerged in the overground trials.Specifically, the DANU sports system exhibited lower reliability during walking and higher reliability during jogging for females in these scenarios.Interestingly, these outcomes did not align with the results obtained in treadmill trials.It's possible that participants needed to deliberately adjust their stride length during overground trials to interact with the force plates, for instance, with females potentially increasing their stride length during walking.This intentional alteration of stride length may have introduced greater variability into their gait patterns (Challis 2001).Similarly, the external prompt of the treadmill at set speeds of walking may reduce the variability of step timing and require less cognitive resources, therefore agreement between sessions would be better for treadmill rather than overground walking at a self-selected pace (Thumm et al 2018, Keller Xin Yu 2021).Therefore, the moderate to excellent test re-test reliability of the DANU system for walking outcomes may reflect taskdependent and intrinsic human variability.

Validity and reliability of gait outcomes during running
During running all reported gait outcomes showed excellent validity (ICC(2,1) >0.90) across overground and treadmill trials, as well as speeds, except for GCT during overground jogging that displayed good accuracy (ICC (2, 1) 0.78) (tables 1 and 2, figures 2 and 3).Validity results of the present study are comparable to previous research using pressure insole devices, where Pearson correlations of 0.84-0.96(Stöggl and Martiner 2017) and 0.99 (Seiberl et al 2018) have been reported.However, within the current study GCT was shorter compared to these previous studies, but was similar to research examining running performance (de Ruiter et al 2014).Minor differences were demonstrated between the DANU and reference systems regarding GCT and swing time.Specifically, the DANU sports system slightly underestimated GCT and over-estimated swing time compared to a gold-standard reference, these differences became more apparent at faster speeds (i.e. 12 and 14 km hr −1 ), especially for GCT.Previous research has shown similar degradation in accuracy of GCT measurement with respect to speed (Falbriard et al 2018, Young et al 2022), highlighting similar limitations in underlying algorithms used across wearable devices.Calculation of GCT is more challenging than other gait outcomes, as it requires more gait event information, such as timings of heel strike and toe off, as well as the orientation and trajectory of the foot (Schuna et al 2013).This is highlighted by the excellent accuracy (ICC(2, 1) >0.90) in stride time, which indicates subtle algorithm differences even between outcome measures of the same system (Donath et al 2016).The DANU sports system calculates temporal gait outcomes through the use of capacitive sensors embedded within a sock form factor, with event detection from a foot-shoe interaction, whereas the laboratory reference system examines a foot-floor interaction.The subtle difference in foot impact event detection between DANU and reference systems may impact event detection accuracy or timing comparison.Similarly, differences in agreement between the wearable and reference systems may be attributed to extraneous noise encountered at higher impact speeds that may be difficult to filter leading misidentification of gait events within underlying algorithms (Young et al 2022).
The test re-test reliability results indicated moderate to excellent agreement for running gait outcomes across the two sessions.Specifically, moderate reliability (ICC(2, 1) 0.67-0.69)was demonstrated for overground running trials, but there was excellent treadmill reliability during treadmill running trials (ICC(2, 1) 0.96-0.99),which was similar to our walking results and was likely affected by the same issues.The reliability results obtained here are comparable to previous research.For example, research examining the Myotest found test-retest reliability for GCT to be poor to moderate at different running speeds (Gouttebarge et al 2015).In contrast to the present study, GCT reliability decreased at slower speeds and lower GCTs were recorded which may relate to testing on an outdoor athletic track, using verbal feedback speeds (de Ruiter et al 2014, Gouttebarge et al 2015), rather than our indoor overground and treadmill laboratory assessment.

Limitations and future directions
Several limitations of the present research and future directions should be highlighted.Firstly, the sample consisted of healthy recreational runners, which may not adequately represent all cohorts of potential interest, such as professional athletes or clinical populations.It is necessary for future studies to evaluate the analytical and clinical validity of the DANU system within specific cohorts of interest to ensure accurate measurement of gait outcomes.Additionally, future studies should determine whether the DANU derived gait metrics are clinically meaningful outcomes through comparison to health metrics (e.g.quality of life, fatigue) or within different sub-populations (e.g.athlete performance level or disease) (Goldsack et al 2020).Furthermore, examination of the usability of the DANU sports system is required within specific populations to ensure generalizability.
Secondly, validity testing was conducted within a laboratory environment using force plates and a treadmill were used during walking and running, which may not entirely represent gait of prolonged overground walking and running in natural environments.Previous research has demonstrated that significant speed by surface interactions exist for the temporal outcomes (Hong et al 2012, Hollis et al 2021).Future work is needed to validate the DANU sports system in more ecologically valid settings (i.e.real-world, community, home environments), as an advantage of wearable devices is the portability and potential to use during real-world prolonged tasks (Benson et al 2018, Meyer et al 2021).
Lastly, data processing of the treadmill trials was conducted as an average over the 60 s trials, whereas the absolute difference of each individual step is reported for overground trials due to the nature of the intermittent overground protocol.Algorithms for gait analysis perform best when processing data from continuous trials, which may be more representative of real-world walking or running than intermittent trials (Ao et al 2018, Seiberl et al 2018, William et al 2021, Straczkiewicz et al 2023).However, future work is needed to compare continuous and intermittent gait outcomes collected in different populations in order to determine the impact on underlying DANU proprietary gait algorithms.

Conclusions
This study examined the analytical validity and test re-test reliability of a commercial wearable technology, the DANU sports system, for measurement of walking and running gait in healthy adults.The DANU system had good to excellent agreement with 3D motion capture in quantifying ground contact time, swing time and stride time during overground and treadmill walking and running at various speeds.Furthermore, the DANU system gait outcomes had moderate to excellent reproducibility across two different sessions in healthy adults.Future research is needed to establish the clinical validity and usability of the DANU Sport System wearable technology to measure gait in various specific populations before routine deployment within performance or clinical settings.
Foundation through a post-doctoral fellowship for basic scientists and a clinical research award (PF-FBS-1898, PF-CRA-2073).

Ethical statement
Ethical approval was granted by the Northumbria University Research Ethics Committee (Reference: 33358) and this study conformed to the Declaration of Helsinki.All participants were supplied with informed consent and gave verbal and written consent before performing testing in Northumbria University's Gait and Biomechanics Laboratory, City Campus.

Figure 1 .
Figure 1.DANU sports system.Reproduced with permission from Danu Sports Limited.

Figure 2 .
Figure 2. Bland-Altman plots illustrating the absolute agreement between the DANU sports system and 3D motion capture system during overground walking and running.(A) Ground contact time (ms), (B) Swing time (ms), (C) Stride Time (ms).

Figure 3 .
Figure 3. Bland-Altman plots illustrating the absolute agreement between the DANU sports system and 3D motion capture system during treadmill walking and running.(A) Ground contact time (ms), (B) Swing time (ms), (C) Stride time (ms).

Rachel
Mason https:/ /orcid.org/0000-0002-5087-1969Gillian Barry https:/ /orcid.org/0000-0003-4886-920XBen O'Callaghan https:/ /orcid.org/0000-0002-5550-6901Alan Godfrey https:/ /orcid.org/0000-0003-4049-9291Samuel Stuart https:/ /orcid.org/0000-0001-6846-9372Tao W et al 2012 Gait analysis using wearable sensors Sensors (Basel) 12 2255-83 Thumm P C et al 2018 Treadmill walking reduces pre-frontal activation in patients with Parkinson's disease Gait Posture 62 384-7 Toth L P et al 2023 Effects of brief intermittent walking bouts on step count accuracy of wearable devices J. Meas.Phys.Behav. 2 13-21 Van Hooren B et al 2020 Is motorized treadmill running biomechanically comparable to overground running? a systematic review and metaanalysis of cross-over studies Sports Med.50 785-813 Veerubhotla A et al 2021 Comparing accuracy of step count detection algorithm at different activity monitor wear locations Arch.Phys.Med.Rehabil.102 e79 Vitale C et al 2012 Effect of global postural rehabilitation program on spatiotemporal gait parameters of parkinsonian patients: a threedimensional motion analysis study Neurological Sci.33 1337-43 Washabaugh E P et al 2017 Validity and repeatability of inertial measurement units for measuring gait parameters Gait.Posture 55 87-93 Johnston W et al 2021 Recommendations for determining the validity of consumer wearable and smartphone step count: expert statement and checklist of the INTERLIVE network Br.J. Sports Med.55 780 Willy R W 2018 Innovations and pitfalls in the use of wearable devices in the prevention and rehabilitation of running related injuries Phys.Therapy Sport 29 26-33 Yamato T P, Saragiotto Bt Fau -Lopes A D and Lopes A D 2015 A consensus definition of running-related injury in recreational runners: a modified Delphi approach American journal of physical medicine & rehabilitation 45 375-80 Young F et al 2022 Examination of a foot mounted IMU-based methodology for a running gait assessment Front.Sports Active Living 4 Zago M et al 2018 Gait analysis in patients with chronic obstructive pulmonary disease: a systematic review Gait Posture 61 408-15 Zhou Y et al 2020 Classification of neurological patients to identify fallers based on spatial-temporal gait characteristics measured by a wearable device Sensors 20 15

Table 1 .
Mean difference, ICC(2,1), limits of agreement (LOA%), and pearson correlation between the reference system and the DANU sports system.